Research And Implementation Of The Scheduling Problem In Cloud Environment

Since cloud computingis proposed, it is rapidly becoming a research hotspot in the field of IT. Cloud computing has brought us the new calculation and service model.In this mode, the user can on-demand access to applications and data on the cloud.As the most basic service in cloud, Infrastructure as a service(IaaS)shares physical resources by virtualization technology. Virtualization allows you to run multiple independent virtual machines on a physical server, andtransparent movement of workloads from one server to another. With the expanding cloud service and the expanding user amount, it is bound to bring greater demand for resources,resources allocation and scheduling algorithm will directly affect the overall system resource utilization, the Quality of Service(QoS) and energy costs.For resource scheduling algorithmin the data cente, academics has a lot of research and achievement. But most studies only foucus on a single object, for example resource utilization, no comprehensive considering multiple aspects, for example the energy costs. In order to improve the resource utilization and reduce the energy consumption of data centres, after in-depth study of existing research results, we propose virtual machineplacement algorithm based on grouping genetic and virtual machine migration policy based on load forecasting.Virtual machine placement algorithm based on grouping geneticis to have better performance in follow terms: resource utilization, energy consumption and peak temperature. In this thesis, after analyzed the disadvantages of genetic algorithm in solving such problems, we put forward improvement group genetic algorithm and apply it to solve virtual machine placement problem.Monitoring the load on each physical machine and using BP neural network to predicte loadof next time period, through two levels of threshold mechanism, virtual machine migration policy based on load forecastingcan effectivelyreduce unnecessary virtual machine migration, enhance the Service-Level Agreement. While selectingvirtual machineneed to be migrated, thealgorithm select VMswith maximummigration efficiency which calculated by load and data.At last, this thesisdesigned a basic cloud computing platform based on OpenStack. In the process of implementation, resource allocation andscheduling algorithmsproposed in this thesisare used.At the same time, we implement resource management middlewares and monitoring system. The results show that the above algorithms are valid and feasible.